Plasma heating with neutral beam injection

Transcription

1 University in Ljubljana Faculty of mathematics and physics Department of physics Plasma heating with neutral beam injection SEMINAR Author: Jure Maglica Mentor: a/prof. dr. Milan Čerček Ljubljana, May 005

2 Abstract Seminar describes how neutral beams for plasma heating are producted. Many areas such as ion production and extraction, acceleration, beam neutralisation and propagation towards tokamak plasma are examined. Because of high energy neutrals demands, positive ion based neutral beam injection (PI-NBI) used recently, is no longer interesting. Negative ion based NBI systems are now in development and are expected to reach 1MeV particle energy and 10MW heating power for 1000s pulses. NBI is very important for tokamak fusion reactors, such as ITER, because it provides plasma heating, current drive and fuel injection. Contents Introduction 3 Neutral beam injection 4 Negative ion neutral beam injection 5 Negative ion production 6 Surface production 6 Volume production 7 Negative ion extraction and acceleration 9 Beam neutralisation 11 Gas neutralis ation 11 Plasma neutralisation 1 Beam transport 13 Neutral beam penetration 13 Plasma heating 15 Achievements 16 Conclusion 17 References 17

3 Introduction Neutral beam heating was developed in the early 1970's, and is now the most important for plasma heating in almost all fusion experiments. Neutral atoms can enter the confining magnetic field magnetic bottle and are ionized in the plasma through collisions with electrons and ions. Magnetic bottle confines fast ions (ionised neutrals) which deliver their energy to plasma electrons and ions that are slower through collisions. It means they heat the plasma (speed up ions) and drive the current (speed up electrons). Fast neutral atom beams are generated with charge exchange neutralization of high energy ion beams. NI-NBI is much more efficient in producing high energy neutral beams than PI-NBI because of neutralisation efficiency at energies higher than 100 kev. For the next generation tokamaks, like ITER, a beam energy of 1- MeV is demanded, and this is a major reason why NI- NBI is being developed, although there is a much bigger problem in producing negative ion beams than positive. In ITER, 50 MW, 1 MeV D beams will be injected in the plasma using three injectors. NI-NBI system is a beam source consisting of a negative ion source, an accelerator and neutralisator. The beam source has to produce a 40A (00 A/m ), 1 MeV D- beam for 1000 s for the NI-NBI system in ITER. [5] Figure 1 : Fusion reactor ITER [14] 3

4 Neutral beam injection Because of magnetic bottle (configuration of magnetic field) there is no possibility to directly inject energetic charged particles inside the plasma. This is a reason for developing fast neutrals. The ions are produced in the plasma source. After extraction they are accelerated to a high energy before crossing a charge exchange cell where they are neutralised. The neutralisation is only partial and the remaining ions are deflected magnetically and sent to a dump. The neutrals can cross the magnetic field and reach the plasma where they get ionised, transferring their energy to the plasma bulk by collisions with electrons and ions. Most NBI are based on positive ion (H+,D+,...) technology. There exist neutral beam injectors able to deliver 1- MW heating power of neutrals at energies up to 150 kev. Plasma source generates various ion species. After extraction by electrostatic potential, the negative ions are eliminated but the molecular ions (D, D3) remain present in the beam. After acceleration, all ions have the same energy, but because molecular ions are heavier, the beam of neutrals delivered to the plasma will contain energies per nucleon of one half and one third (smaller velocity). 30% of the total beam power can be carried by these less energetic components that deposit their energy more at the edge of the plasma compared to D neutrals. This must be taken into account for computing power deposition profiles. Figure : Scheme of tokamak plasma heating [16] When ions cross the neutralisation cell, each one has a neutralisation probability that increases with the length of the neutralisation gas cell. Because neutralised ions can be reionised again, the neutralisation efficiency decreases. Neutralisation and reionisation processes are in equilibrium, based on rate of both cross sections. Each ion has a maximum neutralisation probability for a given lenght of the cell, different for each type of ion. At energies higher than 100keV, neutralisation efficiency for positive ions is lower than 50% and falls under 10% at energies of 300keV. While negative ions have efficiency almost constant of 65% through all energies between 100keV and MeV. This is because of the cross section for losing electron through collisions. The energy that can be reached at reasonable efficiency with positive ion based technology is insufficient for the next generation of machines. For the heating of the ITER plasma at least 0.5MeV beams are required. If the beams are to be 4

5 used to non-inductively generate the plasma current, energies of 1-MeV are required. This is clearly out of reach of any positive-ion based neutral beam and efforts are presently devoted to the development of neutral beams based on a negative-ion source. Major difference between ion beams required for NBI and usual accelerator beams is the much higher beam current required for NBI. If we want to develop the required power 10 MW per beam line, currents around 40 A are necessary. Figure 3 : Schematic diagram of a neutral beam heating system [13] Negative ion neutral beam injection NI-NBI is shematically equal to PI-NBI. But there are some principal differences between them, because the ion source must produce negative ions which is harder to do and NI-NBI operate at much higher energy, around MeV. The electron captured in the negative ion has a very low binding energy called affinity of E a = 0.75 ev. Ion can easily lose the electron and become neutral, and this is a reason why a high neutralisation efficiency can be achieved with negative ions. In order to increase their production of negative ions, caesium is added into the source. It has a very low ionisation potential E i =3.894 ev, and can relatively easy lose electrons which are bound to D. Negative ion sources contain extraction systems, that disables electrons to be accelerated. This is done by putting the field of permanent magnets to the extracting electrostatic field. The stray electrons hit the extractor grid while the negative ion trajectories are nearly unaffected due to higher ion mass. Ions are then accelerated electrostatically up to energies of the order of the MeV and neutralised in gas cell. Two basic types of accelerators are being developed. The MAMuG (for Multi-Aperture Multi-Gap) accelerates ions in steps of around 00 kev. Second type provides the acceleration over a single gap. For all ion beams, the simplest neutralisation cell is a box filled with gas. At energies around 1MeV, the maximum efficiency of gas neutraliser is about 60% for negative ions. Plasma neutralisers can reach an efficiency of up to 85% if the plasma in the cell is fully ionised. Besides stripping of electron another mechanism is happening in a gas cell, that is reionisation of fast neutrals which produces positive ions which are unusable. Both mechanisms happen due to collisions. In gas cell collisions with neutrals and molecules are dominant and in plasma electron and ion collisions are responsible for electron stripping. Because cross section for plasma collisions is bigger than for neutrals, shorter neutralisation cell is needed and bigger efficiency is obtained using plasma neutralisers. 5

6 Figure 4 : Extraction geometry for a three-electrode system for three circular beamlets with offset focussing. [15] Negative ion production So far there are two ways how to make D ions, they are surface and volume production. In surface production the ions are produced when D atoms bounce off walls coated with caesium, which serves as source of electrons. To get a large negative ion yield, very intense wall bombardment is required. It is difficult to operate these sources for long pulses and to produce well focalised ion beams. Volume production is based on a process called dissociative attachment. When D molecule is in a high vibrational state, it breaks up and at the same time it captures an electron. Ion yield is limited, because of the high gas pressure required which leads to neutralisation of the negative ions. Todays most efficient sources (yielding from ma/cm ) of negative ions work through caesium seeding of volume sources, which increases the negative ion yield and minimises the stray electron current. Surface production As mentioned above surface production is one type how ot produce negative ions. The probability of electron capture during backscattering from surface of hydrogen atoms or hydrogen ions depends on the work function Φ of the surface, on the affinity of the electron A, which is 0.75 ev for negative hydrogen ions, and also on the perpendicular velocity of the escaping ion. Φ A γ ( H, H) = exp( Cv ) [13] In equation above C is a constant, and Φ-A represents the difference of energy between the Fermi level of the surface and the electron affinity level. Of course it should be small enough to enhance the biggest electron capture probability. So far the optimum thickness of cesium covering is about 0.6 monolayer (3.3x10 14 Cs/cm ). The best method to lower the surface work function Φ is to cover a metal with cesium. It allows ions of ratio up to 0.67 H - per incident H atom to escape. For tungsten W(110) the work function is reduced from 5,5 ev (pure) to 1.45 ev. In general we could say that negative ions are made in reactions H + e - = H - and H + + e - = H - 6

7 Sources optimized for negative ion surface production should have the following characteristics. Atomic or ionic bombardment of surface should be as intense as possible. The more atoms are in contact with surface, the more negative ions can be produced. The positive ion energy is in the tens of ev range to maximize the negative ion yield and the path of the negative ions in the source is minimized to avoid destruction (losing electron). Path that is long enough can lead to significant neutralisation of already produced D- ions. Volume production In voume production we try to maximize the production of negative ions by dissociative attachment of electrons on high vibrational states of hydrogen molecules. Reaction is H + e - = H + H -. This type of sources was developed, after the surprisingly high production rates of negative ions measured in a hydrogen plasma was found. This is explained by the enhancement of the dissociative attachment cross-sections. When the hydrogen molecules are in high vibrational states, they dissociate and attach an electron. Figure 5 : Scheme of RF source for negative ion production [10] Volume sources consist of two regions separated by a magnetic filter, the driver and the extraction chamber. In first there are arc filaments which emit the primary electrons. Arc voltages are in the order of 100 V between the cathode emitter and the source walls. H gas is injected into this region. Fast electrons excite H molecules with collisions (T e about 5-10 ev) to high vibrational levels. Drift of the excited H molecules to the second chamber where the lower electronic temperature (T e about 1- ev) maximizes the dissociative attachment reaction rates and minimizes the neutralization probability of H - by electronic and ionic collisions. 7

8 Figure 6 : Magnetic field contour map inside ion source [13] There are some significant problems or at least limitations that need to be mentioned. Stray electron currents are produced, and due to high operating pressure we get stripping losses (negative ion neutralisation in acceleration system). Cesium is also used in this type of source although benefits are not well known if they are due to volume or surface production. Cesium decreases surface work function in surface production and lowers the electron temperature in volume production. Both are favorable for greater ion production. Cesium can also change the recombination coefficient, which is directly related to the atomic density. Benefits from cesium seeding are: The negative ion yield is increased by a factor of 3-5 The operating pressure is decreased (less neutralisation) The stray electron current ratio drops from in pure sources to 1-5 The isotopic effect is reduced from D - IH - being about 0.5 in pure sources to about 0.8 Processes that make vibrationally excited molecules H Low energy electron excitation (e-v). Low energy electrons (<5 ev) collide with H molecules and excite them to higher vibrational states. e H 1 * 1 ( X +, v '') e H ( X + + +, v '' + v '') g g For significant population of H (v'') spectrum high collision rate is required. e-v 15 excitation can excite molecules to v'' = 4. Reaction rate for v''=1 is σ v *10. ev For v'' > 1 reaction rate decreases significantly. High energy electron excitation (E-V) This process populates higher vibrational states v'' > 4 and is the most important fo for H - ion volume production. Electrons with high energies (>0eV) excite ground state molecules to higher states, including states that radiative decay into high 1 * 1 vibrational ground state. e + H ( X +, v '' = 0) e + H ( X +, v '' + v '') + hυ Reaction rate for v'' = 7 is σ v are created by this process. g 17 5*10 EV g. Vibrational levels from v'' = 5 and higher 8

9 Figure 7: Potential energy curves for H, H + and H - molecules including some vibrational states. [13] Surface assisted excitation process (s-v) This process creates all possible vibrationally excited molecules H between v'' = 0 and 14. If positive hydrogen ions hit the wall, can get neutralised. * H + * + wall H + e+ wall H wall H + H + e+ wall Reaction rate is proportional to ion velocity and effective surface area of the wall. σ v v A * ion Surface recombination process H atoms can combine with adsorbed H atoms on the surface and form H excited * molecules. Vibrational levels under v'' = 3 are populated. H + H H Creation of H - ions Dissociative attachment (DA) This is the main loss mechanism of excited molecules in the driver region, but is also the main process to produce H - ions in extraction region. Low energy electrons combine with molecules and form ions. States above v'' >5 produce 90% H - ions. ( 1 + e H, '') 15 + X v H + H. Reaction rate for this process is σ v 5*10. High Rydberg states g These states are created through decay of super excited states of H + ion. Lifetime of HR states are 0.1 miliseconds. Required electron temperature to form HR states is above 50 ev. ** * H + + e H ( SES) H ( HR) H + H + vol sur DA 9

10 Figure 8 : Above, schematic diagram of two different types of volume negative ion sources. Below, Profiles of plasma temperature (T e ) and electron (n e ), positive ion (n + ) and negative ion (n - ) density along the axis. [13] Negative ion extraction and acceleration Large current is required for fusion for high power efficiency. It can only be achieved with acceleration in electric field. For positive ion beams, this technique has been developed up to 160 kev. For ITER 1 MeV beams are required, which carry the difficulties expected with high voltages. The problematic of negative ion extraction and acceleration has many features which make it much more complex than that of positive ions. The negatively charged ions are accompanied by co-accelerated stray electrons originating mainly from the source plasma (both are negatively charged), and the fragile negative ions may be neutralized (affinity 0,75 ev) or positively charged, during the acceleration. Figure 9 : Schematic drawing of grid system [11] A beamlet channel of a negative ion extractor and pre-accelerator is shown in the figure. The final beam is composed of many hundreds such beamlets, the extractor and pre-accelerator consisting in a large number of similar channels. Figure 10 : Examples of electron (blue) and ion (red) trajectories for the negative ion extraction grid. [15] Grids are polarized at various voltages, to achieve the desired beam energy. The first grid separates the source plasma from the accelerator. In second grid, often named extraction grid, permanent magnets are inserted to deflect the stray electron currents and to prevent them from being further accelerated, because both negative ions and electrons could be electrostatically accelerated. Magnetic field must be compensated 10

11 by opposite fields, so ion beams do not deflect. For behaviour predictions and simulations, programs are used to numerically solve dynamics and Poisson equations. ρ V = x x ε and m = e( V + B) 0 t t Some electron leakage still occured and secondary emission of electrons and backscattering must be taken into account. Probability for backscattering for electrons with energies around 5eV can be up to 30% which brings stray currents. It means that magnetic configuration must be appropriate. Stripping losses (neutralisation and further ionisaton) are very important to be minimised because they degrade overall performance of the system. These losses can amount up to 0%. They happen due to collisions and can be lower if we reduce operating pressure. Two accelerating sistems are being developed. A multi-hole multi-grid system with long and narrow acceleration channels from the source to the last acceleration stage and second one is merged beam concept in which beamlets are pre-accelerated in a multi-hole system to 100 kev, and then merged into a single beam and accelerated to the full energy in a single gap. Neutralisation Negative ion neutralization is relatively easy due to the low affinity of the additional electron of only 0.75 ev. Neutraliser cell should be designed to provide as high neutralization efficiency as possible, this is why negative ions are being developed instead of positive. Cells should operate with the lowest possible gas input, in order to limit the stripping losses of the negative ion beam in the accelerator and reionization losses of the neutral beam. Figure 11 : Neutralisation efficiencies of different H/D ions. [13] 11

12 Figure 1: Neutralisation Fraction (blue) and required target thickness for 95% neutralisation (red) for D+ ions [11] Gas neutralisation All NI-NBI are based on the use of gas neutralizers. Because outlet beam always consists of negative, neutral and positive atoms, charged particles must be sent to a dump. Neutralisation is based on colisions between fast negative ions and atoms in gas cell. Reaction is H - + H = H + H + e -. When fast ion becomes neutral it can lose another electron. H + H = H + + H + e -. Positive ion is useless for plasma heating. Overall neutralisation efficiency is a functions of cross sections for these two reactions. Because cross section is almost constant for particle energies higher than 10keV, neutralisation efficiency is not a function of beam energy. σ 0 r = 1 1/ 1 η () r = σ 0+ r Figure 13: Cross sections for hydrogen neutralisation (red) and ionisation (black). At high energies neutralisation cross section is very small, but ionisation cross section is almost constant. [5] Plasma neutralisation The plasma neutralizers are based on collisions with charged particles in plasma, where neutralization efficiencies reach up to 85% for fully ionized hydrogen plasma. H - + X(e, Ar, H +, H + ) = H + X + e -.This advantage is increased by the fact that the required target thickness is reduced compared to the gas neutralizer as result of 1

13 greater cross sections. The use of multiple charged Argon or Xenon would enable operating at even lower target thicknesses. Plasma neutralizers must work at high ionization rate, otherwise the collisions with the neutral gas are dominant and the optimum efficiency is lower. For hydrogen, for efficiencies greater than 80% ionization rate of at least 0% is required. Plasma neutralizers will require a lot of additional power (MW), which reduces the gain in the neutralizer efficiency. Electromagnetic stripping I will mention two more possibilities that could be used for electron stripping. This technique has advantage of the fact that a strong electric field gives a finite thickness to the potential barrier for second electron. Tunelling probability increases rapidly with increasing electric field. Mean lifetime t of H - ion in a strong perpendicular / electric field is proportional to t e E b. Because of very high electric fields (MV/cm) needed, this cannot be used for fusion applications Photo deatachment H - + hν = H + e - This reaction has a cross section 4*10-17 cm at hν at around 1,5 ev (λ= 1µm). Photons of such energies are emitted from NdYag lasers. Unfortunately power of such laser system required would be in MW range (continuous regime). Power can be expressed as P= (1 R) hcwvlog(1/1 η) /( ησλ), where η is laser efficiency, R reflection coefficient of laser walls, photo deattachment cross section σ, λ laser wavelength, laser width w and negative ion velocity v. Efficiencies fo up to 100% can be achieved, but additional power needed (gas cell needs no power) decreases overall efficiency. Beam transport Ions that go through neutralisation cell are neutralised only at some efficiencies. Not neutralized ions in the beam may represent a considerable amount of power. Since these ions would be deflected in the magnetic field of the tokamak in an uncontrolled manner. Consequently they would hit surfaces of the system near the entrance aperture and cause significant damage due to thermal overload. Ions in neutral beam systems must be removed in controlled way and to thermalize their power on a suitable ion dump. Ion removal can be done by electrostatic or by magnetic fields. Magnetic field rotates charged particles, while neutrals go through unaffected. It can be seen in figure 3. Field distribution must be well chosen so that thermal loads in dump do not exceed limiting values. With neutrals and ions, cold gas is also released from neutralisation cell and it must be pumped away. Because gas particles are very slow, through collisions they disturb neutral beam. This effect can go even further, to complete beam blocking, if the surfaces hit by the re-ionized particles release more gas. A pressure of less than 10-4 mbar must be maintained in the injection system. Large area cryocondensation pumps (L He, 4 K) or titanium getter pumps (300 K) are being used. Neutral beam penetration Neutral beams are usually injected close to the plasma in parallel direction because it provides the beam the longest path through the densest part of the plasma. As 13

14 plasma is curculating in toroidal direction, beams are usually injected either nearly parallel or nearly perpendicular to it. Perpendicular solution is technologically easiest but the path through the plasma is small. Heated ions are created with large perpendicular energies. Due to magnetic field reflection a substantial fraction of them can be immediately trapped into banana orbits, which can lead to significantly larger losses than using parallel injection. Because of the limited amount of space available in between the toroidal field coils, parallel injection beam lines are harder to design. But their advantage is that they provide a much longer path for the ionisation of the beam. Two sorts of losses are involved in the energy transfer from the neutral beam to the plasma. Shinethrough losses are when some neutrals cross the plasma without being ionised and are lost on the wall opposite to the injection point. Charge exchange losses occur when fast ions get neutralised shortly after their ionisation. The so created neutrals will either leave the plasma or be reionised at an radius nearer to the wall. This leads to direct losses and broadening of the power deposition profile. The neutralisation process is due to charge exchange. The ionisation of the beam is due to several processes such as ionisation by impact on electrons and ions and charge exchange. The dominant process for the lower energy range E=80 kev is charge exchange with slow plasma particles, for higher is ion impact (including impurities) and electron impact, where fast neutrals lose electron. Figure 14: Ionisation cross sections Ionisation rate per length unit is a function of all possible collisions, which can be written as next equation. First two represent ionisation due to electron and proton collisions, third term is for charge exchange with plasma particles and last term for ionisation due to impurities. Of course I is beam intensity and x path length along beams trajectory. di = In ( eσ e + npσ p + npσcx + nzσz) dx collisional ionization by electrons H 0 + e - = H + + e - collisional ionization by plasma ions H 0 + H + = H + + H + + e - charge exchange with plasma ions H 0 + H + = H + + H 0 14

15 If we introduce a beam stopping cross section as np nz nz σ = σe + n ( σ ) e p + σcx + n σ e z σe + σ p + σcx + n σ e z 1 we can define mean free path for beam particles in a plasma with density n λ = nσ. Beam stopping cross section can increase up to 00% if we take into account multi step ionisation. Beam intensity is Il () = Iexp 0 l σ ( r) n( r) dl Plasma heating Plasma heating is due to collisions between beam fast neutrals and plasma electrons and ions. Beam paticles with high energies (1MeV) transfer energy through electron collisions. At some point when equal amount of energy is transferred to both ions and electrons is called critical beam energy. It is shown in next figure and next equation. 0 Figure 15 : Critical energies for hydrogen and deuterium [15] W Mb nizi = 14,8 T ( ) e i c e n M i Simplyfied equation for energy transfer (tokamak plasma heating) is dw 3 b Wb Wc P = = (1 + ( ) ) where first term in brackets represents electron heating dt ts Wb and second ion heating. T s is slowing down time. When W c =W b an equal amount of power is distributed to both species. When beam energy is much larger than critical, equation is simplyfied to dwb W b =, and solution is exponential decay. For example, for α particle plasma dt ts heating W c W α which means that mostly electrons are heated. If we want to compute total energy distribution to both kind of species, electrons and ions, we must dwb calculate P = and integrate W i = Pdt i dt. The result is a fraction of energy distributed to particles 1 W W b0 b 0 3 Wc F = ( ). As mentioned for slow beam particles this i Wb 15

16 fraction is big for ions and low for electrons, the opposite is for fast beam particles, looking against critical velocity. F i + F e = 1. Figure 16: Fractions of energy transferred to electrons F e or ions F i [1] Achievements The N-NBI system of JT-60U (KAMABOKO) has a tangential beamline co-directional to the plasma current. It has two negative ion sources located symmetrically with respect to the equatorial plane at an injection angle of.75. Figure 17 : Three beamlines in LHD, Japan [6] The system is designed to deliver 10 MW of beam power at a beam energy of E = 500 kev for 10 seconds. The overall efficiency of the system is 40%. Up to now 5. MW of deuterium beam has been injected into plasmas at 350 kev for 0.7 seconds (4. MW at 360 kev for hydrogen beam). The individual maximum for energy and pulse duration is 400 kev ( MW, 0.35 s) and 1.9 seconds (300 kev, 1. MW). For ITER required power is 40 MW, for 1000s long pulses produced of 1 MeV 40A D beams. 16

17 Figure 18: Beam energies and current densities achieved [6] Conclusion NBI characteristics are progressing and look very perspective to be used for heating in future fusion reactors. NBI will allow tokamaks work for unlimited periods of time (instead of todays upper limit of 1000s). Many problems still need to be solved to be used as alternative parts that increase efficiency and behaviour of NBI heating system, like RF sources, single gap acceleration, plasma neutralisers etc. Because overall efficiency of such heating system is around 40% a lot of energy will be needed for heating, but tokamak reactor will produce much more energy, so that overall efficiency of reactor will be positive. References [1] A. Theodore Forrester, Large ion beams, Wiley interscience publications, 1987 [] John Wesson, Tokamaks, Oxford science publications, 1987 [3] J. Jacquinot, Steady state operation of tokamaks, Euratom CEA, CEA Cadarache, France [4] M. Hanada et al., Develpment of multi-mega watt negative ion sources and accelerators for neutral beam injectors, Japan atomic energy research institute [5] O. Naito et al., Current drive by negative ion based neutral beam injector in JT- 60U, Japan atomic energy research institute, Conf. On contr. Fusion and plasma physics, C, 118, (1998) [6] O. Kaneko et al., Engineering prospects of negative ion based neutral beam injection, Nucl. Fusion, 43, 69 (003) [7] O. Kaneko et al., Status of negative ion based neutral beam injectors in large helical device, National institute for fusion science Oroshi Toki, Japan [8] J. Pamela, The physics of production, acceleration and neutralisation of large negative ion beams, Plasma phys. and contr. Fusion, 37, 35 (1995) [9] ITER Physics basis, Nucl. Fusion 39, 137 [10] G. Delogu et al., The drift source: A negative ion source module for DC multi amperes ion beams, Euratom CEA, CEA Cadarache, France [11] Y. Okumura et al., Advanced negative ion beam technology to improve the system efficiency of neutral beam injectors, Japan atomic energy research institute [1] I.A. Soloshenko et al., Space charge lens for focusing of negative ion beams, Institute for physics of NAS, Ukraine 17

18 [13] Mainak Bandyopaydhyay, Studies of an inductively coupled negative hydrogen ion radio frequency source through simulations and experiments, Max Planck institut for plasma physics, 004 [14] [15] [16] 18